Index

A

arrays, private

in multistreaming, Multistreaming

assign command

accessing layers provided by FFIO libraries, Does the Code Use Formatted I/O?

bypassing system cache, Does the Code Use Direct Access I/O?

COS-blocked format, Does the Code Use Large, Sequential, Unformatted I/O Requests?

for library buffer size, Does the Code Use Small, Sequential, Unformatted I/O Requests?, Does the Code Use Direct Access I/O?

setting buffer size, Does the Code Use Formatted I/O?

specifying file format, Does the Code Use Asynchronous I/O Requests?

specifying library buffer size, Does the Code Use Large, Sequential, Unformatted I/O Requests?

specifying MR layer, Memory-Resident (MR) Files

to avoid cache, Does the Code Use Asynchronous I/O Requests?

to convert to asynchronous I/O, Does the Code Use Large, Sequential, Unformatted I/O Requests?

to invoke FFIO, Memory-Resident (MR) Files

assign, Identifying I/O Intensive Code

asynchronous I/O

converting to, Does the Code Use Large, Sequential, Unformatted I/O Requests?

optimizing, Does the Code Use Asynchronous I/O Requests?

B

bandwidth, Is the Program Memory Bound?, Latency and Bandwidth Issues

Blocking, Does the Code Use Large, Sequential, Unformatted I/O Requests?, Loop Blocking, Cache Pollution Control

Buffers

translation lookaside, Is the Program Memory Bound?

C

C/C++

loopmark listing, Is the Program Processor Bound?

Cache

avoiding, Does the Code Use Asynchronous I/O Requests?

bandwidth, Bandwidth

coherence, Cache Coherency and Consistency

consistency, Cache Coherency and Consistency

D-cache, D-cache and I-cache

E-cache, E-cache

I-cache, D-cache and I-cache

instruction, Cache

line size, Cache Line Size

locality, Cache Localities and Strides

overview, Overview

pollution control, Cache Pollution Control

scalar data, Cache

stride, Cache Localities and Strides

using no_cache_alloc, Optimizing Memory-bound Code

viewing counters, Optimizing Memory-bound Code

Call stack overflow, Sampling/Asynchronous Experiments

Call stack sample, Sampling/Asynchronous Experiments

compiler options, Using More Aggressive Compiler Options

computed-safe vector length, Computed-safe Vector Length Loops

cpu_time, Common Timing Tools

Cray Streaming Directives, Factors that Enhance Vectorization and Multistreaming

CrayPat, CrayPat

pat_report, Generating a Report

pat_build, Instrumenting a Program

environment variables, Setting the Environment and Running the Experiment

module, Loading the CrayPat Module

sampling experiments, Sampling/Asynchronous Experiments

tracing experiments, Tracing Experiments

D

data edit descriptors, Does the Code Use Formatted I/O?

data items in I/O list

minimizing, Does the Code Use Formatted I/O?

Debugging

prerequisites, Overview

using CrayPat, Sampling and Tracing Experiments, Customizing a Report

direct access I/O

definition, Does the Code Use Direct Access I/O?

directives, Using More Aggressive Compiler Options

double buffering, Does the Code Use Large, Sequential, Unformatted I/O Requests?

E

edit descriptors

repeated, Does the Code Use Formatted I/O?

Examples

mprofil report, Is the Program I/O Bound?, Is the Program Processor Bound?

pat_hwpc report, Is the Program Memory Bound?

direct access I/O, Does the Code Use Direct Access I/O?

expanded samp_pc_time report, Is the Program Processor Bound?

F

FFIO

cache tuning, Does the Code Use Asynchronous I/O Requests?

libraries, Does the Code Use Large, Sequential, Unformatted I/O Requests?

using, Identifying I/O Intensive Code

file format

unblocked for asynchronous I/O, Does the Code Use Asynchronous I/O Requests?

unbuffered and unblocked, Does the Code Use Large, Sequential, Unformatted I/O Requests?

files

memory resident, Memory-Resident (MR) Files

finding the source file, Is the Program I/O Bound?

formatted I/O

increasing library buffer sizes, Does the Code Use Formatted I/O?

optimizing, Does the Code Use Formatted I/O?

reducing amount of, Does the Code Use Formatted I/O?

increasing efficiency, Does the Code Use Formatted I/O?

Fortran

IVDEP directive, Vector Update Loops

loopmark listing, Is the Program Processor Bound?

G

gather, Vector Update Loops

H

hardware counters

viewing with pat_hwpc, Is the Program Memory Bound?

Heap overflow, Sampling/Asynchronous Experiments

Heap sample, Sampling/Asynchronous Experiments

I

I/O-bound code

optimizing, Optimizing I/O-bound Code

I/O

optimizing unformatted, Does the Code Use Large, Sequential, Unformatted I/O Requests?, Does the Code Use Small, Sequential, Unformatted I/O Requests?

tracing, Identifying I/O Intensive Code

unbuffered, unblocked format, Does the Code Use Large, Sequential, Unformatted I/O Requests?

using layered I/O, Identifying I/O Intensive Code

, I/O

asynchronous, Does the Code Use Asynchronous I/O Requests?

asynchronous conversion, Does the Code Use Large, Sequential, Unformatted I/O Requests?

changing to unformatted, Does the Code Use Formatted I/O?

increasing request size, Does the Code Use Small, Sequential, Unformatted I/O Requests?

large requests, Does the Code Use Large, Sequential, Unformatted I/O Requests?

minimizing number of data list items, Does the Code Use Formatted I/O?

optimization principles, Optimizing I/O-bound Code

optimizing for small requests, Does the Code Use Small, Sequential, Unformatted I/O Requests?

optimizing formatted, Does the Code Use Formatted I/O?

IEEE

processor support, Vector Processors

L

latency, Is the Program Memory Bound?, Latency and Bandwidth Issues

layered I/O, Identifying I/O Intensive Code

layer

memory resident, Does the Code Use Small, Sequential, Unformatted I/O Requests?

library buffer sizes, Does the Code Use Large, Sequential, Unformatted I/O Requests?

for formatted I/O, Does the Code Use Formatted I/O?

for unformatted I/O, Does the Code Use Small, Sequential, Unformatted I/O Requests?

loopmark listing

C/C++, Is the Program Processor Bound?

Fortran, Is the Program Processor Bound?

Loop

blocking, Loop Blocking

collapse, Loop Collapse

computed-safe vector length, Computed-safe Vector Length Loops

conditionally vectorized, Conditionally Vectorized Loops

dividing iterations among processors, Multistreaming

fully vectorized, Fully Vectorized Loops

fusion, Loop Fusion

interchange, Loop Interchange

outer-loop vectorization, Outer-loop Vectorization

partially vectorized, Partially Vectorized Loops

partitioned, Streamed and Partitioned Loops

reduction, Reduction Loop

short loops, Short Loops

streamed, Streamed and Partitioned Loops

streaming without partitioning, Streaming without Partitioning

unrolling, Loop Unrolling

vector update, Vector Update Loops

M

Memory bound

defined, Is the Program Memory Bound?

memory-resident

files, Memory-Resident (MR) Files

layer, Does the Code Use Small, Sequential, Unformatted I/O Requests?

Memory

bandwidth, Bandwidth

cache, Overview

cache line size, Cache Line Size

cache pollution control, Cache Pollution Control

changing instruction table size, Optimizing Memory-bound Code

changing page size, Optimizing Memory-bound Code

D-cache, D-cache and I-cache

distributed, Distributed Shared Memory

E-cache, E-cache

I-cache, D-cache and I-cache

other pages, Optimizing Memory-bound Code

segment alignment, Optimizing Memory-bound Code

text pages, Optimizing Memory-bound Code

translation lookaside buffer, Is the Program Memory Bound?

using efficient storage for, Memory-Resident (MR) Files

mflop count, Is the Program Processor Bound?

module, Loading the CrayPat Module

mprofil, Sampling/Asynchronous Experiments

MR layer, Does the Code Use Small, Sequential, Unformatted I/O Requests?

MSP, System Overview, Multistreaming

Multistreaming

and vectorization, Factors that Enhance Vectorization and Multistreaming

factors inhibiting, Factors that Inhibit Multistreaming

loopmark listing, Multistreaming

nested loops with, Factors that Inhibit Multistreaming, Multistreaming

on Cray X1 systems, Multistreaming

private arrays, Multistreaming

processor (MSP), System Overview, Multistreaming

streaming and partitioning, Streamed and Partitioned Loops

streaming without partitioning, Streaming without Partitioning

types of codes optimized, Multistreaming

N

nested loops, Factors that Inhibit Multistreaming

multistreaming, Multistreaming

network optimization, I/O

node modules, System Overview

Node

application, System Overview

cache, Cache Coherency and Consistency

flavor, System Overview

operating system, System Overview

support, System Overview

O

Optimization

I/O, I/O

prerequisites, Overview

process flow, Optimization Flowchart

optimizing code

computed safe vector length, Computed-safe Vector Length Loops

conditionally vectorized, Conditionally Vectorized Loops

I/O bound, Optimizing I/O-bound Code

loop blocking, Loop Blocking

loop collapse, Loop Collapse

loop fusion, Loop Fusion

loop interchange, Loop Interchange

loop unrolling, Loop Unrolling

outer-loop vectorization, Outer-loop Vectorization

partially vectorized loops, Partially Vectorized Loops

reduction loop, Reduction Loop

short loop, Short Loops

streaming, Streamed and Partitioned Loops

vectorization inhibitors, Factors that Inhibit Vectorization

P

parallel programming models, System Overview

pattern matching

loopmark listing, Pattern-Matching

pat_build, Instrumenting a Program

instrumenting a program, Is the Program I/O Bound?

pat_hwpc

viewing cache counters, Optimizing Memory-bound Code

example report, Is the Program Memory Bound?

pat_report, Generating a Report

pat, CrayPat

Performance counter overflow, Sampling/Asynchronous Experiments

Performance counter sample, Sampling/Asynchronous Experiments

PRINT statements, Does the Code Use Formatted I/O?

processor

defined, System Overview

profil, Sampling/Asynchronous Experiments

R

RAID striping, Optimizing I/O-bound Code

READ statements, Does the Code Use Formatted I/O?

records

using longer, Does the Code Use Formatted I/O?

reduction loop, Reduction Loop

registers, Vector Processors

Resource usage overflow, Sampling/Asynchronous Experiments

Resource usage sample, Sampling/Asynchronous Experiments

rtc, Common Timing Tools

RTT, Optimizing Memory-bound Code

S

safe vector length, Computed-safe Vector Length Loops

sampling experiments, Sampling/Asynchronous Experiments

samp_cs_ovfl, Sampling/Asynchronous Experiments

samp_cs_time, Sampling/Asynchronous Experiments

samp_heap_ovfl, Sampling/Asynchronous Experiments

samp_heap_time, Sampling/Asynchronous Experiments

samp_pc_ovfl, Sampling/Asynchronous Experiments

samp_pc_time, Sampling/Asynchronous Experiments

samp_ru_ovfl, Sampling/Asynchronous Experiments

samp_ru_time, Sampling/Asynchronous Experiments

scalar processing, Vectorization

Scalar

data cache, Cache

processor, System Overview

registers, Vector Processors

scatter, Vector Update Loops

scratch files, Minimizing System Calls

secondr, Common Timing Tools

sequential access, Does the Code Use Large, Sequential, Unformatted I/O Requests?

sequential I/O, Does the Code Use Small, Sequential, Unformatted I/O Requests?

optimizing, Does the Code Use Large, Sequential, Unformatted I/O Requests?

single-streaming processors, System Overview

SSP, System Overview

mode availability, System Overview

single-streaming processor, Multistreaming

storage devices

using optimal, Using an Optimal Storage Device

summation loop, Reduction Loop

superscalar processor, System Overview

synchronization reduction, Does the Code Use Asynchronous I/O Requests?

system calls, minimizing, Minimizing System Calls

T

TLB, Is the Program Memory Bound?, Optimizing Memory-bound Code

Tools

cpu_time, Common Timing Tools

pat, CrayPat

rtc, Common Timing Tools

secondr, Common Timing Tools

timef, Common Timing Tools

timex, Common Timing Tools

pat_build, Instrumenting a Program

CrayPat, CrayPat

module, Loading the CrayPat Module

tracing experiments, Tracing Experiments

U

unblocked file format

for asynchronous I/O, Does the Code Use Asynchronous I/O Requests?

unblocked I/O format, Does the Code Use Large, Sequential, Unformatted I/O Requests?

unbuffered I/O format, Does the Code Use Large, Sequential, Unformatted I/O Requests?

unformatted I/O

changing to, Does the Code Use Formatted I/O?

optimizing, Does the Code Use Large, Sequential, Unformatted I/O Requests?, Does the Code Use Small, Sequential, Unformatted I/O Requests?

V

vector processing, Vectorization

Vectorization

and multistreaming, Factors that Enhance Vectorization and Multistreaming

compiler directives, Using More Aggressive Compiler Options

compiler options, Using More Aggressive Compiler Options

computed-safe vector length, Computed-safe Vector Length Loops

conditionally vectorized loops, Conditionally Vectorized Loops

defined, Vectorization

factors inhibiting, Factors that Inhibit Vectorization

fully vectorized loops, Fully Vectorized Loops

loopmark listing, Vectorization

outer loop, Outer-loop Vectorization

partially vectorized loops, Partially Vectorized Loops

reduction loop, Reduction Loop

short loops, Short Loops

types of, Vectorization

vector update loops, Vector Update Loops

Vector

collapse, Reduction Loop

instruction set, Vector Processors

processor, System Overview

registers, Vector Processors

W

workload balancing, Does the Code Use Asynchronous I/O Requests?

WRITE statements, Does the Code Use Formatted I/O?